Skip to main content

Table 4 Quantitative comparison of geometric metrics with state-of-the-art segmentation algorithms

From: Dosimetric impact of deep learning-based CT auto-segmentation on radiation therapy treatment planning for prostate cancer

 

Present work

Balagopal et al. [14]

Sultana et al. [17]

Tong et al. [16]

Prostate

DSC

0.87 ± 0.03

0.90 ± 0.02

0.90 ± 0.05

0.86 ± 0.06

HD\(_{\mathrm{avg}}\)

1.6 ± 0.4

–

1.56 ± 0.37

1.01 ± 0.65

HD\(_{\mathrm{95\%}}\)

4 ± 1

–

5.21 ± 1.2

3.51 ± 1.66

Bladder

DSC

0.96 ± 0.01

0.95 ± 0.02

0.95 ± 0.02

0.96 ± 0.02

HD\(_{\mathrm{avg}}\)

0.95 ± 0.2

–

0.95 ± 0.15

0.97 ± 0.53

HD\(_{\mathrm{95\%}}\)

2.5 ± 0.5

–

4.37 ± 0.56

3.17 ± 3.61

Rectum

DSC

0.89 ± 0.04

0.84 ± 0.04

0.84 ± 0.04

0.86 ± 0.07

HD\(_{\mathrm{avg}}\)

1.4 ± 0.7

–

1.78 ± 1.3

1.22 ± 1.05

HD\(_{\mathrm{95\%}}\)

5 ± 4

–

6.11 ± 1.5

4.34 ± 5.30